← All repos

guillaume-chevalier/Spiking-Neural-Network-SNN-with-PyTorch-where-Backpropagation-engenders-STDP

View on GitHub ↗ ★ 279 Jupyter Notebook

What about coding a Spiking Neural Network using an automatic differentiation framework? In SNNs, there is a time axis and the neural network sees data throughout time, and activation functions are instead spikes that are raised past a certain pre-activation threshold. Pre-activation values constantly fades if neurons aren't excited enough.

4 mentions on Hacker News by 3 people: 1 story, 3 comments · first 2019-08-11 · latest 2023-05-24